[infinispan-issues] [JBoss JIRA] Commented: (ISPN-787) Cache.get(...) does not go remote if DIST + batch + REPEATABLE_READ

[Paul Ferraro (JIRA)]

    [ https://jira.jboss.org/browse/ISPN-787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12564394#comment-12564394 ] 

Paul Ferraro commented on ISPN-787:
-----------------------------------

Digging into this a little further, I think I see where the discrepancy is:

in org.infinispan.container.EntryFactoryImpl.createWrappedEntry(...) returns a NullMarkerEntry for the REPEATABLE_READ case, but a null for the READ_COMMITTED case.

However, org.infinispan.interceptors.DistributionInterceptor.needsRemoteGet(...) will not returns true unless the entry is null, or entry.isLockPlaceholder() is true (for which NullMarkerEntry returns false).

Perhaps entry.isNull() should be added to this criteria?

> Cache.get(...) does not go remote if DIST + batch + REPEATABLE_READ
> -------------------------------------------------------------------
>
>                 Key: ISPN-787
>                 URL: https://jira.jboss.org/browse/ISPN-787
>             Project: Infinispan
>          Issue Type: Bug
>          Components: Distributed Cache
>    Affects Versions: 4.2.0.BETA1
>            Reporter: Paul Ferraro
>            Assignee: Manik Surtani
>            Priority: Critical
>         Attachments: ISPN787TestCase.java
>
>
> The attached test reproduces the problem.
> * Create 3 DIST mode caches from 3 clustered managers
> * Do a cache put
> * If batching is enabled, and isolation level is REPEATABLE_READ, cache get only returns the correct value for those nodes to which the key hashes
> * The same cache get succeeds if isolation level is dropped to READ_COMMITTED or if get is not done within a batch

-- 
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira